Demonstration of High Geospatial Accuracy Achieved by a Fixed Wing UAV
1 (3cm) pixel accuracy across 2 Km grid using a QuestUAV Surveyor Pro UAV
1 Survey Objectives and Deliverable Items
The purpose of this survey was to achieve a high accuracy topographic map and Digital Elevation Model (DEM) of a study area in a foreign country that:
(a) was managed by an independent and competent third party, totally independent to QuestUAV.
(b) provided the opportunity to compare QuestUAV performance against other UAV vendors such as EBEE, Dronematrix, Trimble, UAVER
The survey was conducted in August 2015 in South Korea, under the auspices of LX, the South Korea governmental survey agency.
The agency laid out a set of twelve Ground Control Points (GCP) spread across the 1200m x 1200m survey area.
- Survey report (this document), including a survey description, a summary of the results and a basic image analysis.
- A3 map of the survey area (PDF format).
- Image processing report (Agisoft Photoscan).
- Agisoft Photoscan project file (PSZ format).
- Natural Colour Image with 3cm spatial resolution (GeoTiff, KML and ECW format).
- Digital Elevation Model with 6cm spatial resolution (GeoTiff, KML and ECW format).
Ground Image of the Northern sector of the Survey Area
2 About the Survey, and the Equipment Used
The QuestUAV survey was undertaken on 26 August 2015 with a standard QuestUAV Surveyor Pro (see front page for the UAV) in a built up area close to Jeonjo, Southern Korea. The UAV was equipped with a gimballed Sony A6000 camera operating on a 2 second trigger. 1,350 images were acquired in total with an overlap of 80% in-flight and 60% side lap. The figure below shows the flight path and indicates the image overlap. (The irregular area to the top is the result of high ground reducing the overlap.)
The area of survey has 13 GCP’s and 6 control points for accuracy assessment and is contained within approximately a 1 square kilometre grid.
The UAV took off from the school grounds in the centre of the survey area and flew the area once EAST-WEST and then routed to a NORTH-SOUTH grid, all within a single fight. The UAV returned to the launch area for a parachute landing. The flight took approximately 45 minutes.
A crew of two was used for the survey; a pilot (N King) and a laptop commander (R Moore). The UAV was visible throughout the flight. Flight was autonomous from take off until the decision for parachute landing preparation.
Launching the QuestUAV Surveyor 200 (L) Natural colour image of the study area showing distribution of ground control points (R)
3 Image Processing
An A4 sheet was laid on the ground with an identifying mark in the centre of the sheet. The centre was referenced using high accuracy DGPS survey instruments returning mm accuracy.The image processing was
completed in Agisoft Photoscan. The computer used took approximately 18 hours to complete the dense point cloud creation – the longest part of the processing.
Input for the image processing were the following 3 datasets:
- 1,350 UAV raw images
- QuestUAV log file (image name, latitude, longitude altitude, yaw, pitch, roll)
- 19 ground control points for geo-rectification
A total area of 2.3 square kilometres has been processed. Details on the image processing can be taken from the Agisoft Photoscan Processing Report, which is part of our deliverables.
4 Image Accuracy
The outcome of the image processing was a high resolution Natural Colour mosaic with a spatial resolution of 3cm and a Digital Surface Model (DSM) with a spatial resolution of 6cm.
The accuracy error, calculated through the CP’s, throughout the mosaic was on average one pixel (ie the same as the spatial resolution).
The processed Natural Colour Image shows:
- Land usage: Spread of buildings. Building and land boundaries can be clearly identified. Heights of buildings can be assessed.
- Power lines routing and condition.
- Roads and highways: Sizes, dimensions, road surfaces, barriers, road marking, dangers from signals and signage
- Agricultural information: Field boundaries and field roads can be identified. Crop types, crop status and crop health can be assessed. Presence of illegal crops or illegal usage of land can be detected.
- River boundaries and conditions: Conditions of river structures, flow of water and water colour.
Detail from the Natural Colour Image
Different crop types of agricultural areas can be easily identified.(L) Markings on a road intersection. (R)
4.1 Digital Elevation Model
A Digital Elevation Model (DEM) is a digital representation of the elevation of a terrain. Each pixel of a DEM contains an elevation value. Our
DEM of the study area shows a minimum elevation of 45 meters and a maximum elevation of 105 meters above sea level. The terrain rises from the centre line of the study area in both directions, Northwest and Southeast.
Digital elevation models are the basis for in-depth terrain analysis and hydrologic calculations, like for example:
- Determining the slope of roads
- Calculation of height profiles along roads
- Derivation of contour lines
- Calculation of hill slopes and determination of aspects
- Determining watersheds and stream networks
- Modelling flow accumulation and runoff volumes
Please note there is no universal usage of the terms Digital Elevation Model (DEM), Digital Terrain Model (DTM) and Digital Surface Model (DSM) in scientific literature. In most cases the term digital surface model represents the earth's surface and includes all objects on it. In contrast to a DSM, the digital terrain model represents the bare ground surface without any objects like plants and buildings
In our case we have produced a Digital Surface Model (DSM), showing the elevation of all objects on the ground. A DSM can be used as basis to derive a Digital Terrain Model (DTM).
The accuracy of the DEM was assessed by comparing the elevation values of the ground measurements (GCPs and CPs) with the DEM values at the ground control locations. The table below shows how good the elevation of GCP and DEM match. The average difference is between 0 and 3 cm.
5. Land Mapping and Elevation Analysis
The survey can return a vast array of data in terms of land mapping, infrastructure, elevation change, agriculture, social change, road usage, and many more subjects. The following sub-chapters show examples of an in-depth data analysis.
5.1 Mapping the Location and Size of Buildings
The location and size of buildings or other objects on the ground (field boundaries, ponds, parking areas, etc.) can be precisely determined on the basis of the Natural Colour Image.
Each pixel of the Natural Colour Image has a unique geo-coordinate and represents an area of 3 cm x 3 cm. With such a high spatial resolution, roofs can be easily identified and digitized inside a Geo-Information System (GIS). A GIS allows to automatically calculate the roof area.
Example of mapping greenhouses and determination of roof sizes.
5.2 Determining the Slope of a Road
The slope is a measure of the steepness of a road and can be determined on the basis of a Digital Elevation Model (DEM). Slopes are calculated by determining the change in elevation along two points. Height profiles allow us to understand the elevation changes and show the ups and downs along a track. The figure below shows the height profile along a road section in the north-eastern corner of the study area.
The elevation changes in south-north direction from 49.2m to 53.8m, along a length of 346m. According to the common slope formula (slope = rise/run x 100), the slope of the road section is 1.3 %.
Please visit our dataset page for more examples - Datasets